Run the test above for at least 60 seconds. Here is what each
number means for this game specifically.
Jitter (assists that land out of sync)
A Kameo call is only useful if it arrives in a specific relationship to your own attack. When prediction depth varies from moment to moment, that relationship varies too, and the assist that covered your approach yesterday arrives after it today. This is the most Mortal Kombat 1 specific symptom on the list.
Packet loss (corrections that undo confirmed combos)
Rollback re-simulates frames when the real input differs from the predicted one. With two entities per player there is more state to re-simulate, so a large correction is more visually disruptive here than in a one-character fighter - the opponent and their assist both move.
Ping (and the cross-play pool)
One frame is about 17 ms at sixty frames per second, and rollback conceals a few of them comfortably. Cross-play widens the pool of opponents, which is good for queue times and means a wider spread of connection quality - worth knowing when a session feels inconsistent.
What you see in game, and which number on this page explains it.
The standard account of rollback netcode assumes one character per player. Predict the opponent's input, simulate forward, and if the prediction was wrong, rewind and replay the affected frames. The visible cost of a wrong prediction is that one character ends up somewhere other than where they were drawn, and the eye usually accepts it as movement.
Mortal Kombat 1's Kameo system changes the arithmetic. Each player now drives a fighter and an assist, and the assist's value lies entirely in its timing relative to the fighter's actions - covering an approach, extending a combo, interrupting pressure. That is a second stream of consequential inputs, and a correction does not merely move a character but can change whether two actions that were meant to overlap actually did.
The practical effect is that this game rewards a stable connection more than a fast one, and it does so in a way players find hard to attribute. A late assist does not look like a network problem. It looks like a mistimed input, and the natural response is to adjust the timing - which then breaks offline, and breaks differently on the next connection. Measuring removes that trap.
It is worth saying that the netcode itself is not the weak point here. NetherRealm shipped rollback years before it became the genre's expectation, and the implementation does what it is supposed to do. What it cannot do - what no implementation can - is invent an input that never arrived. That is why the loss and jitter figures on this page are the ones that predict how your evening will go.
If the chart above shows packet loss or high jitter, work through
these in order.
Does Mortal Kombat 1 use rollback netcode?
Yes. NetherRealm adopted rollback well before the genre settled on it as standard, which is why the studio's online play has been comparatively stable for several releases. It hides latency by predicting inputs and correcting when the prediction was wrong.
Why do my Kameo assists arrive late online?
Because an assist is a second controlled entity whose timing has to line up with your own attack, and varying prediction depth changes that relationship frame by frame. Jitter is the figure to look at, not ping.
Should I turn cross-play off?
It depends on what you are optimising for. Cross-play shortens queue times and widens the range of connections you will meet. If your problem is consistency rather than waiting, restricting the pool is a reasonable experiment.
Is packet loss worse than ping in Mortal Kombat 1?
Yes. Rollback is built to conceal a few frames of latency and does it well. Lost packets make the correction later and larger, and with two characters per player there is more state to snap into place when it happens.